Electrical engineering at the University of Michigan. Michigan Athletics creative video intern (Aug 2026–now) and Yazaki North America software engineering co-op (since May 2025).
um-autolog
In development Aug 2026–present
Creative Video Intern / Michigan Athletics
An on-device tool that logs game footage on my Mac, with no cloud. It reads jersey numbers, tags action and audio events like cheering and whistles, and names each clip in ShareBrowser style (e.g. 25BKW_1OSU_005_#05 JUMPER) with a separate detailed description. A review UI lets me check and fix the labels. The naming engine covers 29 sports and 269 action tags, backed by 33 unit tests. It's early and in active development; jersey accuracy is the next focus.
Building a bill of materials from wiring harness design files by hand is slow, repetitive, and error-prone.
Solution
A desktop tool that reads harness design files, builds a structured bill of materials with per-harness and combined views, and exports a multi-sheet Excel workbook, backed by automated regression tests. Built for Tesla EV wiring-harness data, which I owned end to end, from design through deployment and support. Validated 17/17 parts on a production harness and 44/50 (88%) on image-only drawings, vs. a 0-line incumbent baseline.
Checking wiring drawings against taping and clipping rules is manual, and mistakes can slip through to review.
Solution
Compares a wiring drawing against its harness data, flags rule violations right on the drawing with the reason for each, and lets the engineer fix the data and re-validate in the app. Root-caused a systematic defect driving ~$44K in savings per frame harness and cut review time 88% (128 hrs to under 16).